public void AddEnvelopes(int minCharge, int maxCharge, int minScanNum, int maxScanNum,
IList<ObservedIsotopeEnvelope> envelopes = null)
{
var ms1ScanNumToIndex = _run.GetMs1ScanNumToIndex();
var minCol = ms1ScanNumToIndex[minScanNum];
var maxCol = ms1ScanNumToIndex[maxScanNum];
var nRows = maxCharge - minCharge + 1;
var nCols = maxCol - minCol + 1;
MinCharge = minCharge;
MaxCharge = maxCharge;
MinScanNum = minScanNum;
MaxScanNum = maxScanNum;
Envelopes = new ObservedIsotopeEnvelope[nRows][];
for(var i = 0; i < nRows; i++) Envelopes[i] = new ObservedIsotopeEnvelope[nCols];
if (envelopes == null) return;
foreach (var envelope in envelopes)
{
var i = envelope.Charge - MinCharge;
var j = ms1ScanNumToIndex[envelope.ScanNum] - minCol;
if (i < 0 || i >= nRows || j < 0 || j >= nCols) continue;
Envelopes[i][j] = envelope;
}
}
public bool CheckChargeState(ObservedIsotopeEnvelope envelope)
{
var checkCharge = envelope.Charge;
if (checkCharge > 20)
{
return(true); //high charge (> +20), just pass
}
var peakStartIndex = envelope.MinMzPeak.IndexInSpectrum;
var peakEndIndex = envelope.MaxMzPeak.IndexInSpectrum;
var nPeaks = peakEndIndex - peakStartIndex + 1;
if (nPeaks < 10)
{
return(false);
}
if (envelope.NumberOfPeaks > nPeaks * 0.7)
{
return(true);
}
var tolerance = new Tolerance(5);
var threshold = nPeaks * 0.5;
var mzTol = tolerance.GetToleranceAsTh(Spectrum.Peaks[peakStartIndex].Mz);
var minCheckCharge = Math.Max(checkCharge * 2 - 1, 4);
var maxCheckCharge = Math.Min(checkCharge * 5 + 1, 60);
var maxDeltaMz = Constants.C13MinusC12 / minCheckCharge + mzTol;
var nChargeGaps = new int[maxCheckCharge - minCheckCharge + 1];
for (var i = peakStartIndex; i <= peakEndIndex; i++)
{
for (var j = i + 1; j <= peakEndIndex; j++)
{
var deltaMz = Spectrum.Peaks[j].Mz - Spectrum.Peaks[i].Mz;
if (deltaMz > maxDeltaMz)
{
break;
}
for (var c = Math.Round(1 / (deltaMz + mzTol)); c <= Math.Round(1 / (deltaMz - mzTol)); c++)
{
if (c < minCheckCharge || c > maxCheckCharge)
{
continue;
}
var k = (int)c - minCheckCharge;
nChargeGaps[k]++;
if (nChargeGaps[k] + 1 > threshold && nChargeGaps[k] + 1 > 1.25 * envelope.NumberOfPeaks)
{
return(false);
}
}
}
}
return(true);
}
public void UpdateWithDecoyScore(List <Ms1Spectrum> ms1Spectra, int targetMinCharge, int targetMaxCharge)
{
var ms1ScanNumToIndex = _run.GetMs1ScanNumToIndex();
var ms1ScanNums = _run.GetMs1ScanVector();
var minCol = ms1ScanNumToIndex[MinScanNum];
var maxCol = ms1ScanNumToIndex[MaxScanNum];
MinCharge = targetMinCharge;
MaxCharge = targetMaxCharge;
var rnd = new Random();
var comparer = new MzComparerWithBinning(28);
var mostAbuInternalIndex = TheoreticalEnvelope.IndexOrderByRanking[0];
var nRows = MaxCharge - MinCharge + 1;
var nCols = maxCol - minCol + 1;
Envelopes = new ObservedIsotopeEnvelope[nRows][];
for (var i = 0; i < nRows; i++)
{
Envelopes[i] = new ObservedIsotopeEnvelope[nCols];
}
for (var charge = targetMinCharge; charge <= targetMaxCharge; charge++)
{
var mostAbuMz = TheoreticalEnvelope.GetIsotopeMz(charge, mostAbuInternalIndex);
if (_run.MaxMs1Mz < mostAbuMz || mostAbuMz < _run.MinMs1Mz)
{
continue;
}
for (var col = minCol; col <= maxCol; col++)
{
var localWin = ms1Spectra[col].GetLocalMzWindow(mostAbuMz);
var numMzBins = comparer.GetBinNumber(localWin.MaxMz) - comparer.GetBinNumber(localWin.MinMz) + 1;
var peakSet = new Ms1Peak[TheoreticalEnvelope.Size];
for (var k = 0; k < peakSet.Length; k++)
{
var r = rnd.Next(0, numMzBins);
if (r < localWin.PeakCount)
{
peakSet[k] = (Ms1Peak)ms1Spectra[col].Peaks[r + localWin.PeakStartIndex];
}
}
var env = new ObservedIsotopeEnvelope(Mass, charge, ms1ScanNums[col], peakSet, TheoreticalEnvelope);
//AddObservedEnvelope(env);
Envelopes[charge - MinCharge][col - minCol] = env;
}
}
UpdateScore(ms1Spectra, false);
}
internal void Expand(ObservedIsotopeEnvelope envelope)
{
if (MaxScanNum < 0 || envelope.ScanNum > MaxScanNum)
{
MaxScanNum = envelope.ScanNum;
}
if (MinScanNum < 0 || envelope.ScanNum < MinScanNum)
{
MinScanNum = envelope.ScanNum;
}
if (MaxCharge < 0 || envelope.Charge > MaxCharge)
{
MaxCharge = envelope.Charge;
}
if (MinCharge < 0 || envelope.Charge < MinCharge)
{
MinCharge = envelope.Charge;
}
}
public bool CheckChargeState(ObservedIsotopeEnvelope envelope)
{
var checkCharge = envelope.Charge;
if (checkCharge > 20) return true; //high charge (> +20), just pass
var peakStartIndex = envelope.MinMzPeak.IndexInSpectrum;
var peakEndIndex = envelope.MaxMzPeak.IndexInSpectrum;
var nPeaks = peakEndIndex - peakStartIndex + 1;
if (nPeaks < 10) return false;
if (envelope.NumberOfPeaks > nPeaks * 0.7) return true;
var tolerance = new Tolerance(5);
var threshold = nPeaks * 0.5;
var mzTol = tolerance.GetToleranceAsTh(Spectrum.Peaks[peakStartIndex].Mz);
var minCheckCharge = Math.Max(checkCharge * 2 - 1, 4);
var maxCheckCharge = Math.Min(checkCharge * 5 + 1, 60);
var maxDeltaMz = Constants.C13MinusC12 / minCheckCharge + mzTol;
var nChargeGaps = new int[maxCheckCharge - minCheckCharge + 1];
for (var i = peakStartIndex; i <= peakEndIndex; i++)
{
for (var j = i + 1; j <= peakEndIndex; j++)
{
var deltaMz = Spectrum.Peaks[j].Mz - Spectrum.Peaks[i].Mz;
if (deltaMz > maxDeltaMz) break;
for (var c = Math.Round(1 / (deltaMz + mzTol)); c <= Math.Round(1 / (deltaMz - mzTol)); c++)
{
if (c < minCheckCharge || c > maxCheckCharge) continue;
var k = (int)c - minCheckCharge;
nChargeGaps[k]++;
if (nChargeGaps[k] + 1 > threshold && nChargeGaps[k] + 1 > 1.25 * envelope.NumberOfPeaks) return false;
}
}
}
return true;
}
public void AddEnvelopes(int minCharge, int maxCharge, int minScanNum, int maxScanNum,
IList <ObservedIsotopeEnvelope> envelopes = null)
{
var ms1ScanNumToIndex = _run.GetMs1ScanNumToIndex();
var minCol = ms1ScanNumToIndex[minScanNum];
var maxCol = ms1ScanNumToIndex[maxScanNum];
var nRows = maxCharge - minCharge + 1;
var nCols = maxCol - minCol + 1;
MinCharge = minCharge;
MaxCharge = maxCharge;
MinScanNum = minScanNum;
MaxScanNum = maxScanNum;
Envelopes = new ObservedIsotopeEnvelope[nRows][];
for (var i = 0; i < nRows; i++)
{
Envelopes[i] = new ObservedIsotopeEnvelope[nCols];
}
if (envelopes == null)
{
return;
}
foreach (var envelope in envelopes)
{
var i = envelope.Charge - MinCharge;
var j = ms1ScanNumToIndex[envelope.ScanNum] - minCol;
if (i < 0 || i >= nRows || j < 0 || j >= nCols)
{
continue;
}
Envelopes[i][j] = envelope;
}
}
internal void Expand(ObservedIsotopeEnvelope envelope)
{
if (MaxScanNum < 0 || envelope.ScanNum > MaxScanNum)
{
MaxScanNum = envelope.ScanNum;
}
if (MinScanNum < 0 || envelope.ScanNum < MinScanNum)
{
MinScanNum = envelope.ScanNum;
}
if (MaxCharge < 0 || envelope.Charge > MaxCharge)
{
MaxCharge = envelope.Charge;
}
if (MinCharge < 0 || envelope.Charge < MinCharge)
{
MinCharge = envelope.Charge;
}
}
public LcMsPeakCluster(LcMsRun run, ObservedIsotopeEnvelope observedEnvelope)
: this(run, observedEnvelope.TheoreticalEnvelope, observedEnvelope.MonoMass, observedEnvelope.Charge,
observedEnvelope.RepresentativePeak.Mz, observedEnvelope.ScanNum, observedEnvelope.Abundance)
{
}
private LcMsPeakCluster CollectLcMsPeaks(double targetMass, int minRow, int maxRow, int minCol, int maxCol, bool reCollectAllPeaks = false)
{
var ms1ScanNums = Run.GetMs1ScanVector();
var envelopes = new List<ObservedIsotopeEnvelope>();
var bestBcDist = 100d;
ObservedIsotopeEnvelope bestEnvelope = null;
var mostAbuInternalIndex = _theoreticalEnvelope.IndexOrderByRanking[0];
var tolerance = new Tolerance(Comparer.Ppm * 0.5);
var massTol = tolerance.GetToleranceAsTh(targetMass);
var nPeaksCutoff = NumberOfPeaksCutoff;
var bcCutoff = GetSeedBcDistThreshold();
var corrCutoff = GetSeedCorrThreshold();
for (var i = minRow; i <= maxRow; i++)
{
for (var j = minCol; j <= maxCol; j++)
{
if (reCollectAllPeaks) _featureMatrix[i][j].Init();
if (reCollectAllPeaks || !_featureMatrix[i][j].Exist || Math.Abs(_featureMatrix[i][j].AccurateMass - targetMass) > massTol)
{
var peaks = Ms1Spectra[j].GetAllIsotopePeaks(targetMass, i + _targetMinCharge, _theoreticalEnvelope, tolerance);
if (peaks.Count(p => p != null) > 0)
{
_featureMatrix[i][j].DivergenceDist = _theoreticalEnvelope.GetBhattacharyyaDistance(peaks); ;
_featureMatrix[i][j].AccurateMass = targetMass;
_featureMatrix[i][j].CorrelationCoeff = _theoreticalEnvelope.GetPearsonCorrelation(peaks); ;
Array.Copy(peaks, _featureMatrix[i][j].EnvelopePeaks, peaks.Length);
}
}
if (!_featureMatrix[i][j].Exist) continue;
if (_featureMatrix[i][j].CountActivePeaks < nPeaksCutoff) continue;
if (_featureMatrix[i][j].DivergenceDist > bcCutoff && _featureMatrix[i][j].CorrelationCoeff < corrCutoff) continue; // exclude outliers
var envelope = new ObservedIsotopeEnvelope(_featureMatrix[i][j].AccurateMass, i + _targetMinCharge, ms1ScanNums[j], _featureMatrix[i][j].EnvelopePeaks, _theoreticalEnvelope);
envelopes.Add(envelope);
if (_featureMatrix[i][j].EnvelopePeaks[mostAbuInternalIndex] != null && _featureMatrix[i][j].DivergenceDist < bestBcDist)
{
bestBcDist = _featureMatrix[i][j].DivergenceDist;
bestEnvelope = envelope;
}
}
}
if (bestEnvelope == null) return null;
var cluster = new LcMsPeakCluster(Run, bestEnvelope);
cluster.AddEnvelopes(minRow + _targetMinCharge, maxRow + _targetMinCharge, ms1ScanNums[minCol], ms1ScanNums[maxCol], envelopes);
return cluster;
}
private IList<LcMsPeakCluster> GetLcMs1PeakClusters(int binNumber)
{
const int chargeNeighborGap = 4;
var targetMass = Comparer.GetMzAverage(binNumber);
BuildFeatureMatrix(targetMass); // should be called first
var clusters = new List<LcMsPeakCluster>();
// todo : bottom up dataset??
if (_rows.Length < 2 || _cols.Length < 1) return clusters;
var tempEnvelope = new double[_theoreticalEnvelope.Size];
var tempEnvelope2 = new double[_theoreticalEnvelope.Size];
var ms1ScanNums = Run.GetMs1ScanVector();
var ms1ScanNumToIndex = Run.GetMs1ScanNumToIndex();
var mostAbuInternalIndex = _theoreticalEnvelope.IndexOrderByRanking[0];
var tolerance = new Tolerance(Comparer.Ppm*0.5);
foreach (var seed in _seedEnvelopes.OrderBy(s=>s.Key).Select(s=> s.Value))
{
var row = seed.Charge - _targetMinCharge;
var col = ms1ScanNumToIndex[seed.ScanNum];
if (_featureMatrix[row][col].CheckedOutFlag) continue;
var mostAbuMz = _theoreticalEnvelope.GetIsotopeMz(seed.Charge, mostAbuInternalIndex);
var seedLocalWin = Ms1Spectra[col].GetLocalMzWindow(mostAbuMz);
var poissonPvalue = seedLocalWin.GetPoissonTestPvalue(_featureMatrix[row][col].EnvelopePeaks, _theoreticalEnvelope.Size);
var rankSumPvalue = seedLocalWin.GetRankSumTestPvalue(_featureMatrix[row][col].EnvelopePeaks, _theoreticalEnvelope.Size);
var goodEnvelope = (rankSumPvalue < 0.01 || poissonPvalue < 0.01);
if (!goodEnvelope) continue;
var chargeCheck = CorrectChargeState(seed, Ms1Spectra[col]);
if (!chargeCheck) continue;
var seedMass = _featureMatrix[row][col].AccurateMass;
var massTol = tolerance.GetToleranceAsTh(seedMass);
var newCluster = new LcMsPeakCluster(Run, seed);
Array.Clear(tempEnvelope, 0, tempEnvelope.Length);
seed.Peaks.SumEnvelopeTo(tempEnvelope);
var neighbors = new Queue<ObservedIsotopeEnvelope>();
neighbors.Enqueue(seed); // pick a seed
_featureMatrix[row][col].CheckedOutFlag = true;
var summedBcDist = _featureMatrix[row][col].DivergenceDist;
var summedCorr = _featureMatrix[row][col].CorrelationCoeff;
while (neighbors.Count > 0)
{
var cell = neighbors.Dequeue();
var charge = cell.Charge;
var minRw = (int)Math.Max(charge - _targetMinCharge - chargeNeighborGap, _rows.First());
var maxRw = (int)Math.Min(charge - _targetMinCharge + chargeNeighborGap, _rows.Last());
var currCol = ms1ScanNumToIndex[cell.ScanNum];
for (var k = 0; k < 5; k++)
{
var j = currCol;
if (k < 3) j += k;
else j -= (k - 2);
if (j < _cols.First() || j > _cols.Last()) continue;
for (var i = minRw; i <= maxRw; i++)
{
if (_featureMatrix[i][j].CheckedOutFlag) continue;
if (!(_featureMatrix[i][j].AccurateMass > 0)) continue;
if (Math.Abs(seedMass - _featureMatrix[i][j].AccurateMass) > massTol) continue;
Array.Copy(tempEnvelope, tempEnvelope2, tempEnvelope2.Length);
_featureMatrix[i][j].EnvelopePeaks.SumEnvelopeTo(tempEnvelope);
var newDivergence = _theoreticalEnvelope.GetBhattacharyyaDistance(tempEnvelope);
var newCorrelation = _theoreticalEnvelope.GetPearsonCorrelation(tempEnvelope);
if (_featureMatrix[i][j].DivergenceDist < 0.02 ||_featureMatrix[i][j].CorrelationCoeff > 0.7 || newDivergence < summedBcDist || newCorrelation > summedCorr)
{
var envelope = new ObservedIsotopeEnvelope(_featureMatrix[i][j].AccurateMass,
i + _targetMinCharge, ms1ScanNums[j], _featureMatrix[i][j].EnvelopePeaks,
_theoreticalEnvelope);
neighbors.Enqueue(envelope);
newCluster.Expand(envelope);
_featureMatrix[i][j].CheckedOutFlag = true;
summedBcDist = newDivergence;
summedCorr = newCorrelation;
}
else
{
Array.Copy(tempEnvelope2, tempEnvelope, tempEnvelope.Length);
}
}
}
}
LcMsPeakCluster refinedCluster = null;
if (summedCorr > 0.5 || summedBcDist < 0.15)
{
// re-update check-out map
SetCheckOutFlag(newCluster.MinCharge - _targetMinCharge, newCluster.MaxCharge - _targetMinCharge, ms1ScanNumToIndex[newCluster.MinScanNum], ms1ScanNumToIndex[newCluster.MaxScanNum], false);
refinedCluster = GetLcMsPeakCluster(newCluster.RepresentativeMass, newCluster.RepresentativeCharge, newCluster.MinScanNum, newCluster.MaxScanNum, true);
}
if (refinedCluster != null && (_scorer == null || (_scorer != null && refinedCluster.GoodEnougth && refinedCluster.Score >= _scorer.ScoreThreshold)))
{
SetCheckOutFlag(_rows.First(), _rows.Last(), ms1ScanNumToIndex[refinedCluster.MinScanNum], ms1ScanNumToIndex[refinedCluster.MaxScanNum], true);
clusters.Add(refinedCluster);
}
else
{
SetCheckOutFlag(newCluster.MinCharge - _targetMinCharge, newCluster.MaxCharge - _targetMinCharge, ms1ScanNumToIndex[newCluster.MinScanNum], ms1ScanNumToIndex[newCluster.MaxScanNum], true);
}
}
return clusters;
}
//public const double SNRthreshold = 1.4826;
private void BuildFeatureMatrix(double targetMass)
{
InitFeatureMatrix();
SetTargetMass(targetMass);
var observedRows = new BitArray(NRows);
var observedCols = new BitArray(NColumns);
var mostAbuInternalIdx = _theoreticalEnvelope.IndexOrderByRanking[0];
var totalElutionLength = Run.GetElutionTime(Run.MaxLcScan);
var elutionSamplingHalfLen = Math.Max(Math.Min(totalElutionLength * 0.003, 5.0), 0.5);
var neighborHalfColumns = (int) Math.Max((elutionSamplingHalfLen/totalElutionLength)*NColumns, 5);
var targetMassBinNum = Comparer.GetBinNumber(targetMass);
var tolerance = new Tolerance(Comparer.Ppm*0.5);
var minMs1Mz = _ms1PeakList.First().Mz;
var maxMs1Mz = _ms1PeakList.Last().Mz;
var nPeaksCutoff = NumberOfPeaksCutoff;
var bcSeedCutoff = GetSeedBcDistThreshold();
var corrSeedCutoff = GetSeedCorrThreshold();
var ms1ScanNums = Run.GetMs1ScanVector();
var ms1ScanNumToIndex = Run.GetMs1ScanNumToIndex();
var options = new ParallelOptions();
if (_maxThreadCount > 0) options.MaxDegreeOfParallelism = _maxThreadCount;
_seedEnvelopes.Clear();
Parallel.ForEach(_rows, options, row =>
{
var charge = row + _targetMinCharge;
for (var col = 0; col < NColumns; col++) _featureMatrix[row][col].Init();
for (var k = 0; k < _theoreticalEnvelope.Size; k++)
{
var i = _theoreticalEnvelope.IndexOrderByRanking[k];
var isotopeIndex = _theoreticalEnvelope.Isotopes[i].Index;
var isotopeMzLb = (k == 0) ? Ion.GetIsotopeMz(Comparer.GetMzStart(targetMassBinNum), charge, isotopeIndex) : Ion.GetIsotopeMz(Comparer.GetMzAverage(targetMassBinNum - 1), charge, isotopeIndex);
var isotopeMzUb = (k == 0) ? Ion.GetIsotopeMz(Comparer.GetMzEnd(targetMassBinNum), charge, isotopeIndex) : Ion.GetIsotopeMz(Comparer.GetMzAverage(targetMassBinNum + 1), charge, isotopeIndex);
if (isotopeMzLb < minMs1Mz || isotopeMzUb > maxMs1Mz) continue;
var st = _ms1PeakList.BinarySearch(new Ms1Peak(isotopeMzLb, 0, 0));
if (st < 0) st = ~st;
for (var j = st; j < _ms1PeakList.Count; j++)
{
var ms1Peak = _ms1PeakList[j];
if (ms1Peak.Mz > isotopeMzUb) break;
var col = ms1Peak.Ms1SpecIndex;
if (k == 0) // most abundant peak
{
if (_featureMatrix[row][col].EnvelopePeaks[i] == null || ms1Peak.Intensity > _featureMatrix[row][col].EnvelopePeaks[i].Intensity)
{
_featureMatrix[row][col].EnvelopePeaks[i] = ms1Peak;
_featureMatrix[row][col].AccurateMass = Ion.GetMonoIsotopicMass(ms1Peak.Mz, charge, isotopeIndex);
}
}
else
{
if (!(_featureMatrix[row][col].AccurateMass > 0)) continue;
var expectedPeakMz = Ion.GetIsotopeMz(_featureMatrix[row][col].AccurateMass, charge, isotopeIndex);
if (Math.Abs(expectedPeakMz - ms1Peak.Mz) > tolerance.GetToleranceAsTh(ms1Peak.Mz)) continue;
// in case of existing isotope peaks, select peaks maximizing envelope similairty
if (_featureMatrix[row][col].EnvelopePeaks[i] != null)
{
if (_featureMatrix[row][col].CountActivePeaks == 1)
{
if (ms1Peak.Intensity > _featureMatrix[row][col].EnvelopePeaks[i].Intensity) _featureMatrix[row][col].EnvelopePeaks[i] = ms1Peak;
}
else
{
var tmpPeak = _featureMatrix[row][col].EnvelopePeaks[i];
var bc1 = _theoreticalEnvelope.GetBhattacharyyaDistance(_featureMatrix[row][col].EnvelopePeaks);
_featureMatrix[row][col].EnvelopePeaks[i] = ms1Peak;
var bc2 = _theoreticalEnvelope.GetBhattacharyyaDistance(_featureMatrix[row][col].EnvelopePeaks);
if (bc1 < bc2) _featureMatrix[row][col].EnvelopePeaks[i] = tmpPeak;
}
}
else
{
_featureMatrix[row][col].EnvelopePeaks[i] = ms1Peak;
}
}
}
if (k == 0)
{
// for cells missing most abundant peaks
for (var col = 0; col < NColumns; col++)
{
if (_featureMatrix[row][col].Exist) continue;
var highestIntensity = 0d;
var inferredAccurateMass = 0d;
// find the most intense abundant peak from neighboring cells
for (var j = Math.Max(col - neighborHalfColumns, 0); j <= Math.Min(col + neighborHalfColumns, NColumns - 1); j++)
{
var mostAbuPeak = _featureMatrix[row][j].EnvelopePeaks[mostAbuInternalIdx];
if (mostAbuPeak != null && mostAbuPeak.Intensity > highestIntensity)
{
highestIntensity = mostAbuPeak.Intensity;
inferredAccurateMass = _featureMatrix[row][j].AccurateMass;
}
}
_featureMatrix[row][col].AccurateMass = inferredAccurateMass;
}
}
}
for (var col = 0; col < NColumns; col++)
{
if (!(_featureMatrix[row][col].Exist)) continue;
if (_featureMatrix[row][col].CountActivePeaks >= nPeaksCutoff)
{
var corr = _theoreticalEnvelope.GetPearsonCorrelation(_featureMatrix[row][col].EnvelopePeaks);
var bcDist = _theoreticalEnvelope.GetBhattacharyyaDistance(_featureMatrix[row][col].EnvelopePeaks);
_featureMatrix[row][col].CorrelationCoeff = corr;
_featureMatrix[row][col].DivergenceDist = bcDist;
if (!observedRows[row]) observedRows[row] = true;
if (!observedCols[col]) observedCols[col] = true;
// collect seed envelopes
var mostAbuPeak = _featureMatrix[row][col].EnvelopePeaks[mostAbuInternalIdx];
if (mostAbuPeak != null && (bcDist < bcSeedCutoff || corr < corrSeedCutoff))
{
var signalToNoiseRatio = mostAbuPeak.Intensity / Ms1Spectra[col].MedianIntensity;
if (signalToNoiseRatio > 3)
{
var seed = new ObservedIsotopeEnvelope(_featureMatrix[row][col].AccurateMass, row + _targetMinCharge, ms1ScanNums[col], _featureMatrix[row][col].EnvelopePeaks, _theoreticalEnvelope);
lock (_seedEnvelopes)
{
_seedEnvelopes.Add(new KeyValuePair<double, ObservedIsotopeEnvelope>(bcDist, seed));
}
}
}
}
else
{
_featureMatrix[row][col].AccurateMass = 0d;
}
}
}// end or row for-loop
);
var temp = new List<int>();
for (var i = 0; i < observedRows.Length; i++) if (observedRows[i]) temp.Add(i);
_rows = temp.ToArray();
temp.Clear();
for (var i = 0; i < observedCols.Length; i++) if (observedCols[i]) temp.Add(i);
_cols = temp.ToArray();
}
private bool CorrectChargeState(ObservedIsotopeEnvelope envelope, Ms1Spectrum spectrum)
{
if (envelope.Charge > 20) return true; //high charge (> +20), just pass
var peaks = spectrum.Peaks;
var peakStartIndex = envelope.MinMzPeak.IndexInSpectrum;
var peakEndIndex = envelope.MaxMzPeak.IndexInSpectrum;
var intensityThreshold = envelope.HighestIntensity * 0.15;
var nPeaks = 0;
for (var i = peakStartIndex; i <= peakEndIndex; i++)
{
if (peaks[i].Intensity > intensityThreshold) nPeaks++;
}
if (envelope.NumberOfPeaks > nPeaks * 0.7) return true;
//var tolerance = new Tolerance(5);
var tolerance = new Tolerance(Comparer.Ppm * 0.5);
var threshold = nPeaks * 0.5;
var threshold2 = envelope.NumberOfPeaks + (envelope.TheoreticalEnvelope.Size - 1) * 0.7;
var mzTol = tolerance.GetToleranceAsTh(peaks[peakStartIndex].Mz);
var minCheckCharge = Math.Max(envelope.Charge * 2 - 1, 4);
var maxCheckCharge = Math.Min(envelope.Charge * 5 + 1, 60);
var maxDeltaMz = Constants.C13MinusC12 / minCheckCharge + mzTol;
var nChargeGaps = new int[maxCheckCharge - minCheckCharge + 1];
for (var i = peakStartIndex; i <= peakEndIndex; i++)
{
if (!(peaks[i].Intensity > intensityThreshold)) continue;
for (var j = i + 1; j <= peakEndIndex; j++)
{
if (!(peaks[j].Intensity > intensityThreshold)) continue;
var deltaMz = peaks[j].Mz - peaks[i].Mz;
if (deltaMz > maxDeltaMz) break;
if (Math.Abs(deltaMz - mzTol) < float.Epsilon)
{
// Peaks are too close together; continue
continue;
}
for (var c = Math.Round(1 / (deltaMz + mzTol)); c <= Math.Round(1 / (deltaMz - mzTol)); c++)
{
if (c < minCheckCharge)
continue;
if (c > maxCheckCharge)
break;
var k = (int)c - minCheckCharge;
nChargeGaps[k]++;
if (nChargeGaps[k] + 1 > threshold && nChargeGaps[k] + 1 > threshold2) return false;
}
}
}
return true;
}
public IsotopeEnvelopeStatisticalInfo PreformStatisticalSignificanceTest(ObservedIsotopeEnvelope envelope)
{
int peakStartIndex;
Tuple <double, double> mzBoundary;
//var refPeak = envelope.Peaks[envelope.RefIsotopeInternalIndex];
var mostAbuMz = 0d;
var mostAbutPeakInternalIndex = envelope.TheoreticalEnvelope.IndexOrderByRanking[0];
if (envelope.Peaks[mostAbutPeakInternalIndex] != null)
{
mostAbuMz = envelope.Peaks[mostAbutPeakInternalIndex].Mz;
}
else
{
mostAbuMz = envelope.TheoreticalEnvelope.GetIsotopeMz(envelope.Charge, mostAbutPeakInternalIndex);
}
var rankings = GetLocalRankings(mostAbuMz, out peakStartIndex, out mzBoundary);
// smallest delta_mz = 0.01 (th) ?
var ret = new IsotopeEnvelopeStatisticalInfo
{
LocalMzStart = mzBoundary.Item1,
LocalMzEnd = mzBoundary.Item2,
NumberOfLocalPeaks = rankings.Length,
NumberOfPossiblePeaks = (int)Math.Ceiling(100 * (mzBoundary.Item2 - mzBoundary.Item1)),
NumberOfIsotopePeaks = envelope.Size,
};
// calculate ranksum test score
var ranksum = 0;
var nRankSum = 0;
for (var i = 0; i < envelope.Size; i++)
{
if (envelope.Peaks[i] == null || !envelope.Peaks[i].Active)
{
continue;
}
ret.NumberOfMatchedIsotopePeaks++;
//if (isotopeList[i].Ratio > RelativeIntesnityThresholdForRankSum)
//{
var localIndex = envelope.Peaks[i].IndexInSpectrum - peakStartIndex;
if (localIndex >= rankings.Length || localIndex < 0)
{
continue;
}
ranksum += rankings[localIndex];
nRankSum++;
//}
}
var pvalue = FitScoreCalculator.GetRankSumPvalue(ret.NumberOfLocalPeaks, nRankSum, ranksum);
ret.RankSumScore = (pvalue > 0) ? -Math.Log(pvalue, 2) : 50;
// calculate poisson test score
var n = ret.NumberOfPossiblePeaks;
var k = ret.NumberOfIsotopePeaks; // # of theretical isotope ions of the mass within the local window
var n1 = ret.NumberOfLocalPeaks; // # of detected ions within the local window
var k1 = ret.NumberOfMatchedIsotopePeaks; // # of matched ions generating isotope envelope profile
var lambda = ((double)n1 / (double)n) * k;
pvalue = 1 - Poisson.CDF(lambda, k1);
ret.PoissonScore = (pvalue > 0) ? -Math.Log(pvalue, 2) : 50;
return(ret);
}
public void UpdateScore(List <Ms1Spectrum> ms1Spectra, bool pValueCheck = true)
{
var nRows = MaxCharge - MinCharge + 1;
var ms1ScanNumToIndex = _run.GetMs1ScanNumToIndex();
var minCol = ms1ScanNumToIndex[MinScanNum];
var maxCol = ms1ScanNumToIndex[MaxScanNum];
var nCols = maxCol - minCol + 1;
var mostAbuIdx = TheoreticalEnvelope.IndexOrderByRanking[0];
ClearScore();
var bestChargeDist = new double[] { 10.0d, 10.0d };
// sum envelopes at each charge
var summedIntensity = new double[TheoreticalEnvelope.Size];
var xicLen = nCols + 18;
var xicStartIdx = 9;
/*
* if (nCols < 13)
* {
* xicLen = 13;
* xicStartIdx = (int) Math.Floor((xicLen - nCols)*0.5);
* }*/
var xic2 = new double[2][];
xic2[0] = new double[xicLen];
xic2[1] = new double[xicLen];
var chargeXic = new double[nRows][];
var tempBestBcDist = 10.0d;
var repEnvelopeBcDist = 10.0d;
ObservedIsotopeEnvelope repEnvelope = null;
var repEnvelopeBcDist2 = 10.0d;
ObservedIsotopeEnvelope repEnvelope2 = null;
var tempBestDistanceScoreAcrossCharge = new double[2] {
10, 10
};
var tempBestIntensityScoreAcrossCharge = new double[2];
var tempBestCorrelationScoreAcrossCharge = new double[2];
for (var i = 0; i < nRows; i++)
{
var charge = i + MinCharge;
var mostAbuMz = TheoreticalEnvelope.GetIsotopeMz(charge, mostAbuIdx);
Array.Clear(summedIntensity, 0, summedIntensity.Length);
chargeXic[i] = new double[xicLen];
var chargeIdx = (charge % 2 == 0) ? EvenCharge : OddCharge;
var summedMostAbuIsotopeIntensity = 0d;
var summedReferenceIntensity = 0d;
for (var j = 0; j < nCols; j++)
{
var envelope = Envelopes[i][j];
var col = minCol + j;
var localWin = ms1Spectra[col].GetLocalMzWindow(mostAbuMz);
if (envelope == null)
{
continue;
}
envelope.Peaks.SumEnvelopeTo(summedIntensity);
var mostAbuPeak = envelope.Peaks[mostAbuIdx];
if (mostAbuPeak != null && mostAbuPeak.Active)
{
summedMostAbuIsotopeIntensity += mostAbuPeak.Intensity;
summedReferenceIntensity += localWin.HighestIntensity;
}
AbundanceDistributionAcrossCharge[chargeIdx] += envelope.Abundance;
var newBcDist = TheoreticalEnvelope.GetBhattacharyyaDistance(envelope.Peaks);
var newCorr = TheoreticalEnvelope.GetPearsonCorrelation(envelope.Peaks);
var goodEnvelope = (newBcDist <0.07 || newCorr> 0.7);
if (goodEnvelope)
{
xic2[chargeIdx][xicStartIdx + j] += envelope.Abundance;
chargeXic[i][xicStartIdx + j] = envelope.Abundance;
}
var levelOneEnvelope = true;
var levelTwoEnvelope = true;
if (pValueCheck)
{
var poissonPvalue = localWin.GetPoissonTestPvalue(envelope.Peaks, TheoreticalEnvelope.Size);
var rankSumPvalue = localWin.GetRankSumTestPvalue(envelope.Peaks, TheoreticalEnvelope.Size);
levelOneEnvelope = (rankSumPvalue < 0.01 && poissonPvalue < 0.01);
//levelTwoEnvelope = (rankSumPvalue < 0.05 || poissonPvalue < 0.05);
}
if (levelOneEnvelope)
{
if (newBcDist < BestDistanceScoreAcrossCharge[chargeIdx])
{
BestDistanceScoreAcrossCharge[chargeIdx] = newBcDist;
if (localWin.MedianIntensity > 0)
{
BestIntensityScoreAcrossCharge[chargeIdx] = envelope.HighestIntensity / localWin.HighestIntensity;
}
else
{
BestIntensityScoreAcrossCharge[chargeIdx] = 1.0d;
}
}
BestCorrelationScoreAcrossCharge[chargeIdx] = Math.Max(BestCorrelationScoreAcrossCharge[chargeIdx], newCorr);
if (newBcDist < repEnvelopeBcDist)
{
repEnvelopeBcDist = newBcDist;
repEnvelope = envelope;
}
// in the initial scoring, classify major and minor envelopes
if (!_initScore && goodEnvelope)
{
envelope.GoodEnough = true;
}
}
if (levelTwoEnvelope)
{
if (newBcDist < tempBestDistanceScoreAcrossCharge[chargeIdx])
{
tempBestDistanceScoreAcrossCharge[chargeIdx] = newBcDist;
if (localWin.MedianIntensity > 0)
{
tempBestIntensityScoreAcrossCharge[chargeIdx] = envelope.HighestIntensity / localWin.HighestIntensity;
}
else
{
tempBestIntensityScoreAcrossCharge[chargeIdx] = 1.0d;
}
}
tempBestCorrelationScoreAcrossCharge[chargeIdx] = Math.Max(tempBestCorrelationScoreAcrossCharge[chargeIdx], newCorr);
if (newBcDist < repEnvelopeBcDist2)
{
repEnvelopeBcDist2 = newBcDist;
repEnvelope2 = envelope;
}
}
}
var bcDist = TheoreticalEnvelope.GetBhattacharyyaDistance(summedIntensity);
EnvelopeDistanceScoreAcrossCharge[chargeIdx] = Math.Min(bcDist, EnvelopeDistanceScoreAcrossCharge[chargeIdx]);
EnvelopeCorrelationScoreAcrossCharge[chargeIdx] = Math.Max(TheoreticalEnvelope.GetPearsonCorrelation(summedIntensity), EnvelopeCorrelationScoreAcrossCharge[chargeIdx]);
if (BestCharge[chargeIdx] < 1 || bcDist < bestChargeDist[chargeIdx])
{
BestCharge[chargeIdx] = charge;
bestChargeDist[chargeIdx] = bcDist;
if (summedReferenceIntensity > 0)
{
EnvelopeIntensityScoreAcrossCharge[chargeIdx] = summedMostAbuIsotopeIntensity / summedReferenceIntensity;
}
//if (summedMedianIntensity > 0) EnvelopeIntensityScoreAcrossCharge[chargeIdx] = Math.Min(1.0, 0.1*(summedMostAbuIsotopeIntensity / summedMedianIntensity));
}
if (bcDist < tempBestBcDist)
{
tempBestBcDist = bcDist;
Array.Copy(summedIntensity, RepresentativeSummedEnvelop, RepresentativeSummedEnvelop.Length);
}
}
// when good envellope is observed at only either even or odd charge...
if (BestCorrelationScoreAcrossCharge[0] > 0.7 && BestCorrelationScoreAcrossCharge[1] < 0.5)
{
const int i = 1;
BestCorrelationScoreAcrossCharge[i] = tempBestCorrelationScoreAcrossCharge[i];
BestIntensityScoreAcrossCharge[i] = tempBestIntensityScoreAcrossCharge[i];
BestDistanceScoreAcrossCharge[i] = tempBestDistanceScoreAcrossCharge[i];
}
if (BestCorrelationScoreAcrossCharge[1] > 0.7 && BestCorrelationScoreAcrossCharge[0] < 0.5)
{
const int i = 0;
BestCorrelationScoreAcrossCharge[i] = tempBestCorrelationScoreAcrossCharge[i];
BestIntensityScoreAcrossCharge[i] = tempBestIntensityScoreAcrossCharge[i];
BestDistanceScoreAcrossCharge[i] = tempBestDistanceScoreAcrossCharge[i];
}
// normalize abudnace across charges
var s = AbundanceDistributionAcrossCharge[0] + AbundanceDistributionAcrossCharge[1];
if (s > 0)
{
for (var chargeIdx = 0; chargeIdx < 2; chargeIdx++)
{
AbundanceDistributionAcrossCharge[chargeIdx] = AbundanceDistributionAcrossCharge[chargeIdx] / s;
}
}
if (nCols > 1)
{
var evenChargeIdx = BestCharge[EvenCharge] - MinCharge;
var oddChargeIdx = BestCharge[OddCharge] - MinCharge;
XicCorrelationBetweenBestCharges[0] = FitScoreCalculator.GetPearsonCorrelation(Smoother.Smooth(chargeXic[evenChargeIdx]), Smoother.Smooth(chargeXic[oddChargeIdx]));
XicCorrelationBetweenBestCharges[1] = FitScoreCalculator.GetPearsonCorrelation(Smoother.Smooth(xic2[EvenCharge]), Smoother.Smooth(xic2[OddCharge]));
}
if (repEnvelope == null && repEnvelope2 != null)
{
repEnvelope = repEnvelope2;
}
if (repEnvelope != null)
{
// set representative charge, mz and scanNum
RepresentativeCharge = repEnvelope.Charge;
RepresentativeMz = repEnvelope.RepresentativePeak.Mz;
RepresentativeScanNum = repEnvelope.ScanNum;
}
_initScore = true;
}
public LcMsPeakCluster(LcMsRun run, ObservedIsotopeEnvelope observedEnvelope)
: this(run, observedEnvelope.TheoreticalEnvelope, observedEnvelope.MonoMass, observedEnvelope.Charge,
observedEnvelope.RepresentativePeak.Mz, observedEnvelope.ScanNum, observedEnvelope.Abundance)
{
}
public void UpdateWithDecoyScore(List<Ms1Spectrum> ms1Spectra, int targetMinCharge, int targetMaxCharge)
{
var ms1ScanNumToIndex = _run.GetMs1ScanNumToIndex();
var ms1ScanNums = _run.GetMs1ScanVector();
var minCol = ms1ScanNumToIndex[MinScanNum];
var maxCol = ms1ScanNumToIndex[MaxScanNum];
MinCharge = targetMinCharge;
MaxCharge = targetMaxCharge;
var rnd = new Random();
var comparer = new MzComparerWithBinning(28);
var mostAbuInternalIndex = TheoreticalEnvelope.IndexOrderByRanking[0];
var nRows = MaxCharge - MinCharge + 1;
var nCols = maxCol - minCol + 1;
Envelopes = new ObservedIsotopeEnvelope[nRows][];
for (var i = 0; i < nRows; i++) Envelopes[i] = new ObservedIsotopeEnvelope[nCols];
for (var charge = targetMinCharge; charge <= targetMaxCharge; charge++)
{
var mostAbuMz = TheoreticalEnvelope.GetIsotopeMz(charge, mostAbuInternalIndex);
if (_run.MaxMs1Mz < mostAbuMz || mostAbuMz < _run.MinMs1Mz) continue;
for (var col = minCol; col <= maxCol; col++)
{
var localWin = ms1Spectra[col].GetLocalMzWindow(mostAbuMz);
var numMzBins = comparer.GetBinNumber(localWin.MaxMz) - comparer.GetBinNumber(localWin.MinMz) + 1;
var peakSet = new Ms1Peak[TheoreticalEnvelope.Size];
for (var k = 0; k < peakSet.Length; k++)
{
var r = rnd.Next(0, numMzBins);
if (r < localWin.PeakCount)
peakSet[k] = (Ms1Peak) ms1Spectra[col].Peaks[r + localWin.PeakStartIndex];
}
var env = new ObservedIsotopeEnvelope(Mass, charge, ms1ScanNums[col], peakSet, TheoreticalEnvelope);
//AddObservedEnvelope(env);
Envelopes[charge - MinCharge][col - minCol] = env;
}
}
UpdateScore(ms1Spectra, false);
}
public IsotopeEnvelopeStatisticalInfo PreformStatisticalSignificanceTest(ObservedIsotopeEnvelope envelope)
{
int peakStartIndex;
Tuple<double, double> mzBoundary;
//var refPeak = envelope.Peaks[envelope.RefIsotopeInternalIndex];
var mostAbuMz = 0d;
var mostAbutPeakInternalIndex = envelope.TheoreticalEnvelope.IndexOrderByRanking[0];
if (envelope.Peaks[mostAbutPeakInternalIndex] != null)
{
mostAbuMz = envelope.Peaks[mostAbutPeakInternalIndex].Mz;
}
else
{
mostAbuMz = envelope.TheoreticalEnvelope.GetIsotopeMz(envelope.Charge, mostAbutPeakInternalIndex);
}
var rankings = GetLocalRankings(mostAbuMz, out peakStartIndex, out mzBoundary);
// smallest delta_mz = 0.01 (th) ?
var ret = new IsotopeEnvelopeStatisticalInfo
{
LocalMzStart = mzBoundary.Item1,
LocalMzEnd = mzBoundary.Item2,
NumberOfLocalPeaks = rankings.Length,
NumberOfPossiblePeaks = (int)Math.Ceiling(100 * (mzBoundary.Item2 - mzBoundary.Item1)),
NumberOfIsotopePeaks = envelope.Size,
};
// calculate ranksum test score
var ranksum = 0;
var nRankSum = 0;
for (var i = 0; i < envelope.Size; i++)
{
if (envelope.Peaks[i] == null || !envelope.Peaks[i].Active) continue;
ret.NumberOfMatchedIsotopePeaks++;
//if (isotopeList[i].Ratio > RelativeIntesnityThresholdForRankSum)
//{
var localIndex = envelope.Peaks[i].IndexInSpectrum - peakStartIndex;
if (localIndex >= rankings.Length || localIndex < 0) continue;
ranksum += rankings[localIndex];
nRankSum++;
//}
}
var pvalue = FitScoreCalculator.GetRankSumPvalue(ret.NumberOfLocalPeaks, nRankSum, ranksum);
ret.RankSumScore = (pvalue > 0) ? -Math.Log(pvalue, 2) : 50;
// calculate poisson test score
var n = ret.NumberOfPossiblePeaks;
var k = ret.NumberOfIsotopePeaks; // # of theretical isotope ions of the mass within the local window
var n1 = ret.NumberOfLocalPeaks; // # of detected ions within the local window
var k1 = ret.NumberOfMatchedIsotopePeaks; // # of matched ions generating isotope envelope profile
var lambda = ((double)n1 / (double)n) * k;
pvalue = 1 - Poisson.CDF(lambda, k1);
ret.PoissonScore = (pvalue > 0) ? -Math.Log(pvalue, 2) : 50;
return ret;
}